Earth and Planetary Science Letters 273 (2008) 94–104

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

Rapid accretion and differentiation of iron meteorite parent bodies inferred from182Hf–182W chronometry and thermal modeling

Liping Qin a,b,⁎, Nicolas Dauphas a,b, Meenakshi Wadhwa b,1, Jozef Masarik c, Philip E. Janney b,1

a Origins Laboratory, Department of the Geophysical Sciences and Enrico Fermi Institute, The University of Chicago, 5734 South Ellis Avenue, Chicago, IL 60637, USAb Department of Geology, The Field Museum, 1400 South Lake Shore Drive, Chicago, IL 60605, USAc Komensky University, Department of Nuclear Physics, Mlynska dolina F/1, SK-842 15 Bratislava, Slovakia

⁎ Corresponding author. Present address: DepartmCarnegie Institution of Washington, 5241 Broad Branc20015, USA.

E-mail address: lqin@ciw.edu (L. Qin).1 Present address: Center for Meteorite Studies,

Exploration, Arizona State University, Box 871404, Temp

0012-821X/$ – see front matter © 2008 Elsevier B.V. Adoi:10.1016/j.epsl.2008.06.018

A B S T R A C T

A R T I C L E I N F O

Article history:

New high-precision W isoto Received 26 February 2008Received in revised form 3 June 2008Accepted 11 June 2008Available online 25 June 2008

Editor: T. Spohn

Keywords:Hf-W model agemetal-silicate differentiationaccretion timescales

pe measurements are presented for 33 iron meteorites from 8 magmatic groups(IC, IIAB, IID, IIIAB, IIIE, IIIF, I VA and IVB), 2 non-magmatic groups (IAB–IIICD and IIE), and one ungroupediron (Deep Springs). All magmatic irons have ε182W values that are, within errors, equal to, or less radiogenicthan, the Solar System initial of −3.47±0.20. A method was developed to correct the measured ε182W valuesof magmatic iron meteorites for the presence of cosmogenic effects produced during space exposure togalactic cosmic rays. The corrected data provide new constraints on the timing of metal-silicatedifferentiation in iron meteorite parent bodies, which must have taken place within a few million years(b2 to 6 My) of condensation of calcium–aluminum-rich inclusions (CAIs). Metal-silicate differentiation ages(from 182Hf–182W systematics) were combined with parent body sizes (from metallographic cooling rates)into a model of planetesimal heating by 26Al-decay, to constrain the accretion timescale of iron meteoriteparent bodies. Accretion of iron meteorite parent bodies most likely occurred within 1.5 My of the formationof CAIs. The fast accretion times of iron meteorite parent bodies are consistent with dynamical modelsindicating that these objects may have originated in the terrestrial planet-forming region, where theaccretion rates were high. Our W isotopic data for non-magmatic IAB–IIICD and IIE irons provide newconstraints for their formation mechanisms. In particular, they support formation of IAB–IIICD ironmeteorites by melting during a single collision event dated at 4–7 My after formation of the Solar System.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Several extant and extinct radiochronometers (187Re–187Os,107Pd–107Ag, 53Mn–53Cr) have been used to constrain the time offormation of magmatic iron meteorites (Shen et al., 1996; Smoliaret al., 1996; Chen and Wasserburg, 1996; Carlson and Hauri, 2001;Sugiura and Hoshino, 2003). These systems indicate that accretion,differentiation, and crystallization of iron meteorite parent bodiesoccurred early, within the first several tens of My of the formationof the Solar System. However, they all have limitations to theirapplication as fine scale chronometers for early Solar Systemevents. Hafnium-182 is an extinct radionuclide which β− decaysinto 182Ta with a half-life of 8.9 My (Vockenhuber et al., 2004). Thelatter is very unstable (t1/2=115 d) and rapidly decays to 182W. The182Hf–182W system is useful to date the relative timing of metal-

ent of Terrestrial Magnetism,h Road, NW, Washington, DC

School of Earth and Spacee, AZ 85287, USA.

ll rights reserved.

silicate differentiation in the early Solar System (Jacobsen andHarper, 1996; Lee and Halliday, 1996; Horan et al., 1998; Yin et al.,2002; Schoenberg et al., 2002; Kleine et al., 2002; Quitté and Birck,2004; Lee, 2005; Kleine et al., 2005; Scherstén et al., 2006;Markowski et al., 2006a,b) because Hf is lithophile while W ismoderately siderophile. If bulk planetesimals had chondriticcompositions and metal-silicate differentiation occurred aftercomplete decay of 182Hf, metallic cores should have chondritic182W compositions. If differentiation occurred while 182Hf was stillalive, a deficit in 182W in cores relative to chondrites should beproduced. Thus, by studying the isotopic abundance of 182W inmagmatic iron meteorites, which are thought to representfragments of planetesimal cores, we can infer the timing ofmetal-silicate differentiation relative to condensation of refractoryCAIs in the protosolar nebula.

Horan et al. (1998) performed the first extensive study of Wisotopes in iron meteorites. They showed that magmatic ironmeteorites have ε182W values (relative deviation of the ratio of 182Wto a stable W isotope from a terrestrial standard in parts per 104)between −5.1 and −3.1. Recent studies have shown that some ironmeteorites, such as Tlacotepec, have very negative ε182W values (−4.4to −4.0) (Quitté and Birck, 2004; Lee, 2005; Scherstén et al., 2006;

95L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

Markowski et al., 2006a,b), lower than the initial solar value of −3.47±0.20 defined by CAI and chondrite isochrons (Kleine et al., 2005).Taken at face value, the implication would be that metal-silicatedifferentiation in these iron meteorite parent bodies predated theformation of CAIs, which goes against the current paradigm that CAIsare the oldest condensed material in the Solar System.

Exposure to galactic cosmic radiation (GCR) can modify the Wisotopic compositions of iron meteorites (Masarik, 1997; Leya et al.,2003) through capture of secondary thermal neutrons. This can lowerε182W values and yield false negative ages relative to CAIs (Masarik,1997; Leya et al., 2003). Markowski et al. (2006b) measured the Wisotopic compositions along depth profiles in Grant (IIIB) and Carbo(IID) iron meteorites. They demonstrated that the anomalously lowε182W values in these meteorites were produced by interactions withGCR. Thus in those samples, ε182W cannot be interpreted in terms of182Hf decay alone. It is conceivable that the very negative ε182W valuesmeasured in Tlacotepec have also been produced by exposure to GCR,given the long exposure age (945±55Ma) of this sample (Voshage andFeldmann, 1979). A major challenge in establishing the 182Hf–182Wsystem as a reliable chronometer is to find a way to accurately correctfor the cosmogenic effect. Knowledge of exposure ages alone is notsufficient because the effect of irradiation by GCR is modulated by thedepth of burial in the pre-atmospheric object (Masarik, 1997).

We studied Hf–W systematics in iron meteorites using animproved method, allowing variations of less than 0.1 ε-unit on182W to be resolved (Foley et al., 2005; Qin et al., 2007; Qin et al.,2008). Increasing the precision of W isotopic measurements in ironmeteorites can improve the time resolution of core segregationprocesses, and provide a potential means of correcting cosmogenicand nucleosynthetic effects. Several aspects of our approach distin-guish this work from previous studies:

(i) This study focused on a subset of samples with wellcharacterized 41K–40K exposure ages (Voshage and Feldmann,1979; Voshage et al., 1983; Voshage, 1984).

(ii) A precision of ∼±0.1 or better on ε182W was achieved, which iscomparable to or better than themost recent studies (Schersténet al., 2006; Markowski et al., 2006a), but represents a ∼2–5fold improvement compared to earlier efforts (Jacobsen andHarper, 1996; Lee and Halliday, 1996; Horan et al., 1998; Yinet al., 2002; Schoenberg et al., 2002; Kleine et al., 2002; Quittéand Birck, 2004; Lee, 2005).

(iii) A new method is presented for estimating ε182W of magmaticiron meteorites prior to GCR exposure.

(iv) The time of core–mantle differentiation, as inferred from 182Hf–182W systematics, was used in a thermal model of planetesimaldifferentiation to constrain the time of accretion of ironmeteorite parent bodies.

2. Methods

The methods used for purifying W and analyzing its isotopiccomposition have been presented in detail elsewhere (Foley et al.,2005; Qin et al., 2007) and will be briefly reviewed here. Ironmeteorite samples were first leached in 11 N HCl–1 N HF to removesurface-sited terrestrial contamination from various sources. Thecleaned samples were then dissolved in aqua regia and evaporated todryness and redissolved in 11 N HCl. Chemical separation wasachieved by passing the samples through one cation exchange columnand several anion-exchange columns (Qin et al., 2007). This protocolcan accommodate sample sizes up to 1.7 g so that large quantities ofclean W (usually 500–2000 ng) can be retrieved for isotopic analyses.The final W solutions were analyzed on a Micromass Isoprobe multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS)located at the Field Museum, Chicago. The measurements werenormalized to 186W/183W=1.98594 (Völkening et al., 1991) using the

exponential law. This law describes well the fractionation occurring ina MC-ICPMS (Maréchal et al., 1999). The NIST 3163 W referencematerial was used as the W isotope standard. Both the samples andstandards were run at ion intensities of 3–8×10−11 A for 184W,obtained with 30–80 ppb solutions. The sample and standardsolutions were introduced into the ICP source in the form of aerosolsusing an Aridus desolvating nebulizer. The W concentration of thesample was always matched with that of the bracketing standardwithin 3% becausewe noticed that a mismatch in theW concentrationcan affect the accuracy of the isotope measurements (Qin et al., 2007).The sample measurements were interspersed between those of thestandard, and the internally normalizedW isotope composition of thesample was corrected using the mean of the two bracketing standardanalyses. All isotope ratios are presented using the ε notation, alwaysusing 183W as the denominator isotope. A total of 13–20 (n) repeatswere obtained for each sample and were used to calculate averagesand 95% confidence intervals as follows,

ɛ ¼ 1n∑n

k¼1ɛk F

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1

n−1∑n

k¼1ɛk−ɛð Þ2

st0:95;n−1ffiffiffi

np ð1Þ

where t0.95,n−1 is student's t-value corresponding to a two-sided 95%confidence interval for n−1 degrees of freedom.

The absolute 182W/183Wand 184W/183Wratios (after normalization to186W/183W=1.98594) of the NIST 3163 W reference material averagedover a period of two years are 1.85174 (±43, 2σ) and 2.14123 (±63, 2σ),respectively. These values agree within error with those reportedpreviously by negative thermal ionizationmass spectrometry (1.85128±35 and 2.14078±13) and MC-ICPMS (1.85163±8 and 2.14076±7)(Völkening et al., 1991; Lee and Halliday, 1995).

3. Results

3.1. ε182W results

Iron meteorites from both magmatic (IC, IIAB, IID, IIIAB, IIIE, IIIF, I VAand IVB) and non-magmatic (IAB–IIICD and IIE) groups were studied(Table 1). All samples show deficits in ε182Wof −4.2 to −2.3 relative to theterrestrialW reference NIST 3163, a proxy for the composition of the bulksilicate Earth (Fig. 1). These values are also significantly lower than thosemeasured inbulk chondrites (−1.9±0.2) (Yin et al., 2002; Schoenberget al.,2002; Kleine et al., 2004). This range of variation is consistent withpreviouswork, anddemonstrates that 182Hfwas alive at the timeofmetal-silicate differentiation in thesemeteorites (Jacobsen andHarper,1996; Leeand Halliday, 1996; Horan et al., 1998; Yin et al., 2002; Schoenberg et al.,2002; Kleine et al., 2002; Quitté and Birck, 2004; Lee, 2005; Kleine et al.,2005; Scherstén et al., 2006;Markowski et al., 2006a). Resolvable isotopicvariations are present both within and between iron meteorite groups.

Some of the samples measured in this study have also beenanalyzed in previous work, including Nocoleche (IC) (Scherstén et al.,2006), Arispe (IC) (Lee, 2005; Scherstén et al., 2006; Markowski et al.,2006a), Bendego (IC) (Scherstén et al., 2006), El Burro (IIA) (Schersténet al., 2006), Carbo (IID) (Markowski et al., 2006a,b), Henbury (IIIAB)(Horan et al., 1998; Scherstén et al., 2006; Markowski et al., 2006a),Grant (IIIB) (Lee, 2005; Markowski et al., 2006b), Nelson County (IIIF)(Markowski et al., 2006a), Duchesne (IVA) (Scherstén et al., 2006),Tawallah Valley (IVB) (Horan et al., 1998; Markowski et al., 2006a),Cape of Good Hope (IVB) (Horan et al., 1998; Lee, 2005; Schersténet al., 2006; Markowski et al., 2006a), Santa Clara (IVB) (Markowskiet al., 2006a), Hoba (IVB) (Scherstén et al., 2006), Tlacotepec (IVB)(Horan et al., 1998; Quitté and Birck, 2004; Lee, 2005; Kleine et al.,2005; Scherstén et al., 2006; Markowski et al., 2006a), and Watson(IIE) (Snyder et al., 2001; Markowski et al., 2006a). There is goodagreement with the present work except for Tlacotepec (Kleine et al.,2005) and Nocoleche (Scherstén et al., 2006) that can be explained bydifferences in the degree of shielding from GCR.

Table 1Measured W isotopic compositions of iron meteorites

Group Sample Exposure⁎⁎ age Pre-atmospheric mass ε182W ε184W Model age N

(Ma) (103 kg) (My)

Magmatic IC Nocoleche 250±70 −3.58±0.05 0.03±0.04 −0.8±1.7 13Arispe 955±90 ∞ (2π) −3.91±0.06 0.04±0.03 −3.1±1.7 18Bendego 940±90 ∞ (2π) −3.97±0.06 0.08±0.04 −3.5±1.7 16

IIAB Smithsonian⁎ 90±80 −3.42±0.05 0.02±0.06 0.4±1.7 15Sierra Gorda⁎ 140±110 −3.47±0.08 −0.02±0.07 0.0±1.8 13Cedartown #1a⁎ 180±80 −3.36±0.05 0.01±0.04 0.9±1.7 17Cedartown #1b⁎ 180±80 −3.33±0.18 0.04±0.10 1.2±2.3 5Cedartown #2⁎ 180±80 −3.58±0.10 0.00±0.11 −0.9±1.8 9Sample mean −3.39±0.04 0.01±0.03 0.6±1.7El Burro⁎ 165±115 −3.52±0.11 −0.01±0.09 −0.4±1.9 16

IID Carbo 850±140 5 −4.09±0.08 −0.03±0.04 −4.3±1.7 17IIIAB Ruff's Mountain 120 −3.53±0.06 0.04±0.04 −0.5±1.7 15

Henbury 240 −3.70±0.07 0.01±0.03 −1.8±1.7 15Sacramento Mountains 315±55 2 −3.46±0.06 −0.02±0.07 0.1±1.7 14Grant #1(G 01) 695±65 2 −3.54±0.09 0.01±0.08 −0.5±1.7 15Grant #2 (G05) 695±65 2 −3.66±0.07 −0.01±0.05 −1.5±1.7 14Grant #3 (G10) 695±65 2 −3.69±0.13 0.01±0.05 −1.7±1.9 14Grant #4 (G10) 695±65 2 −3.56±0.12 0.03±0.06 −0.7±1.9 13

IIIE Rhine Villa 325±70 5 −3.58±0.09 0.02±0.04 −0.9±1.8 14Kokstad 470±90 ∞ (2π) −3.65±0.06 −0.02±0.06 −1.4±1.7 15

IIIF Nelson County 490±55 2 −3.30±0.05 −0.05±0.03 1.5±1.7 17IVA Gibeon #1⁎ 32 −3.36±0.09 0.04±0.07 1.0±1.8 12

Gibeon #2⁎ 32 −3.39±0.06 0.07±0.05 0.7±1.7 14Gibeon #3⁎ 32 −3.41±0.15 0.02±0.06 0.5±2.1 15Sample mean −3.38±0.05 0.05±0.04Duchesne 220±70 −3.33±0.27 0.07±0.14 1.2±2.9 12Bishop Canyon 220 −3.56±0.11 0.07±0.03 −0.7±1.8 15Wood's Mountain 280 −3.36±0.14 −0.03±0.07 0.9±2.1 14Hill City 475±90 3 −3.48±0.18 −0.05±0.09 −0.1±2.2 12

IVB Tawallah Valley #1⁎ 250±85 −3.53±0.06 −0.10±0.02 −0.2±1.7 18Tawallah Valley #2⁎ 250±85 −3.53±0.05 −0.07±0.05 −0.1±1.7 16Sample mean −3.53±0.04 −0.10±0.02Santa Clara⁎ −3.62±0.06 −0.06±0.03 −0.9±1.7 16Hoba⁎ 340±110 100 −3.43±0.08 −0.07±0.05 0.7±1.8 16Cape of Good Hope⁎ 775±70 30 −3.71±0.06 −0.12±0.04 −1.5±1.7 15Skookum⁎ 945±90 1 −3.55±0.1 −0.11±0.05 −0.3±1.8 16Tlacotepec #1a⁎ 945±55 3 −4.02±0.07 −0.10±0.06 −3.6±1.7 17Tlacotepec #1b⁎ 945±55 3 −3.95±0.04 −0.05±0.02 −3.2±1.7 7Tlacotepec #2⁎ 945±55 3 −4.04±0.09 −0.09±0.04 −3.7±1.7 14Tlacotepec #3⁎ 945±55 3 −4.14±0.07 −0.06±0.05 −4.3±1.7 17Tlacotepec #4⁎ 945±55 3 −4.22±0.08 −0.07±0.06 −4.8±1.7 13Tlacotepec #5⁎ 945±55 3 −4.22±0.07 −0.11±0.03 −4.8±1.7 15Tlacotepec #6⁎ 945±55 3 −4.25±0.05 −0.09±0.02 −4.9±1.7 16Tlacotepec #7⁎ 945±55 3 −4.11±0.16 −0.05±0.08 −4.2±1.9 11Tlacotepec #8a⁎ 945±55 3 −4.20±0.07 −0.09±0.05 −4.6±1.7 17Tlacotepec #8b⁎ 945±55 3 −4.13±0.06 −0.05±0.04 −4.2±1.7 15Tlacotepec #9⁎ 945±55 3 −4.01±0.05 −0.15±0.04 −3.5±1.7 25Sample mean −4.10±0.02 −0.08±0.01

Non-magmatic IAB–IIICD BoHumilitz 140±230 −3.02±0.09 0.06±0.07 4.3±2.0 11Surprise Springs 130±170 −2.94±0.07 0.04±0.04 5.3±1.9 16Deport 1140±70 50 −3.82±0.05 0.05±0.04 −2.6±1.7 17Nantan −2.93±0.07 0.01±0.05 5.4±1.9 15

IIE Watson 8 −2.30±0.08 0.02±0.04 17.5±3.9 15Arlington 150 −3.01±0.08 0.03±0.04 4.5±1.9 15

Ungrouped Deep Springs #1a⁎ 2275±65 5 −3.80±0.06 −0.19±0.05 −1.9±1.7 16Deep Springs #1b⁎ 2275±65 5 −3.75±0.06 −0.14±0.03 −1.6±1.7 2Deep Springs #2⁎ 2275±65 5 −3.78±0.05 −0.15±0.06 −1.8±1.7 13Deep Springs #3⁎ 2275±65 5 −3.77±0.10 −0.26±0.09 −1.8±1.8 12Deep Springs #4⁎ 2275±65 5 −3.74±0.11 −0.12±0.05 −1.5±1.8 10Sample mean −3.77±0.02 −0.15±0.02

ε182W and ε184W are the relative deviations from NIST 3163 W standard; εiW=((iW/183W)sample/(iW/183W)std−1)×104. Pre-atmospheric masses are from Voshage (1984).N represents the number of repeats. Arabic numeral suffixes represent separate dissolutions of different pieces of the same meteorite sample. Small letter suffixes represent differentaliquots of the same dissolution.⁎Results from our previous work (Qin et al., 2007; Qin et al., 2008).⁎⁎Exposure ages from Signer and Nier (1962); Schultz (1967); Nord and Zahringer (1972); Voshage and Feldmann (1979); Villa et al. (1981); Voshage et al. (1983); Voshage (1984);Olsen et al. (1994); Nishiizumi et al. (1995); Lavielle et al. (1997); Lavielle et al. (1999).

96 L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

All magmatic irons have ε182W values that are, within error, equalto or less radiogenic than the Solar System initial value (−3.47±0.20)derived from a CAI isochron (Kleine et al., 2005). This result isconsistent with previous studies. In contrast to magmatic irons, non-magmatic irons are characterized by more radiogenic ε182W values

ranging from −3.82±0.05 to −2.30±0.08, with amajority at −3.00. Thisis also consistent with previous work (Horan et al., 1998; Schersténet al., 2006; Markowski et al., 2006). One IAB–IIICD iron meteorite,Deport, has a very unradiogenic ε182W of −3.82±0.05 (Table 1; Fig. 1),consistent with its long GCR exposure age of 1140 Ma.

Fig. 2. A) Sample locations of three specimens taken from within a slice of Grant. Theslice had been cut into a number of bars (labeled A to T). The distances of the specimensG01, G05 and G10 from the reference line (vertical line) are 325, 200 and 84 mm,respectively. The black dot shows the location of the pre-atmospheric center, accordingto (Hoffman and Nier,1958). B) and C) Plots of measured ε182W values vs. distance to thepre-atmospheric center for Grant and Carbo. Also shown are modeled profiles for thetwo meteorites, with three different libraries: ENDF, JEF and KASKAD (see Kollar et al.,2006 for details). The circles and squares represent the values obtained in this study andin (Markowski et al., 2006b), respectively. The exposure ages for Grant and Carbo are695 and 850 My (Voshage and Feldmann, 1979).

Fig. 1. ε182W values measured in iron meteorites. Circles and triangles correspond tomagmatic and non-magmatic iron meteorites, respectively. Non-magmatic groups areindicated with asterisks. Solid symbols and open symbols represent iron meteoriteswith long exposure ages (N600 My) and relatively short exposure ages (b600 My)respectively; the 600 My exposure age cut-off is arbitrary. The error bars represent 95%confidence intervals.

97L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

Three pieces of the Grant, a group IIIB iron meteorite, wereanalyzed (Table 1). This meteorite has an exposure age of 695 Ma(Voshage and Feldmann, 1979). The 3 specimens were sampled from asection of Grant that has beenwell characterized in terms of its cosmicray exposure (see Fig. 2A for sample locations in the section and(Hoffman and Nier, 1958; Signer and Nier, 1960) for noble gas data). Ascan be seen in Fig. 2B, our results are comparable with those ofMarkowski et al. (2006b). A general positive correlation is found, i.e.,the ε182W values increase with increasing distance from the center (ordecreasing depth from pre-atmospheric surface), although all the datapoints are within error of each other. Note that Markowski et al.(2006b) adopted a pre-atmospheric center based on a recent noble gasstudy (Ammon et al., 2006), which differs significantly from thatsuggested in previous work (Hoffman and Nier, 1958; Signer and Nier,1960). In Fig. 2B, the distances in (Markowski et al., 2006b) wererecalculated using the pre-atmospheric center determined in (Hoff-man and Nier, 1958; Signer and Nier, 1960).

3.2. ε184W results

The ε184W results are shown in Fig. 3. The majority of the samplesshow no deviation in ε184W from the terrestrial NIST W standardoutside of the analytical uncertainty (∼±0.1 ε unit). In particular, thetwo largest magmatic iron groups, IIIAB and IIAB have ε184W values of0 within error bars. This is consistent with a recent study byMarkowski et al. (2006a). However, all IVB iron meteorites and the

ungrouped iron Deep Springs show systematic deficiencies in ε184Wof −0.08±0.01 and −0.15±0.02, respectively. The deficiencies in ε184Win IVBs have been replicated in different laboratories using 3 different

Fig. 3. ε184W values measured in iron meteorites. The symbols are the same as in Fig. 1.See Qin et al. (2008) for the significance of negative ε184W values in some ironmeteoritegroups.

Fig. 4.Measured ε182W values of magmatic iron meteorites vs. their exposure ages. Thethree lines show the model results (maximum cosmogenic effect, i.e., optimum radiusand depth) computedwith the three libraries and assuming a pre-exposure ε182W valueof −3.47.

98 L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

instruments (Isoprobe at the Field Museum in Chicago, Nu-plasma atETH in Zurich, and Neptune at the Thermo Factory in Bremen) and inthe case of IVBs, may reflect incomplete mixing of products of s- and r-process nucleosynthesis in the solar nebula (Qin et al., 2008). Qin et al.(2008) estimated the correction that needs to be applied to ε182Wmodel ages due to this possible nucleosynthetic effect. The conclusionis that the age correction is small (b0.5 My; see Qin et al. 2008 fordetails). Note that several other samples may also show non-zeroε184W values (e.g. Bishop Canyon and Deport), but systematic testsand replicates on these samples were not done to assess thesignificance of these results.

4. Discussion

As has already been pointed out in a number of studies, a majordifficulty with 182Hf-182W chronology of iron meteorites is that somespecimens yield model ages that pre-date CAIs. Assuming a simpletwo-stage evolution model, in which only one Hf/W fractionationevent associated with core formation occurs, the ε182W values of ironmeteorites can be used to calculate metal-silicate differentiation ages,

ΔTA−B ¼ 1λln

ɛ182WA−ɛ182WC

ɛ182WB−ɛ182WC

� �; ð2Þ

where λ is the decay constant of 182Hf (0.078±0.002 My−1)(Vockenhuber et al., 2004), ε182WC is the present day value ofcarbonaceous chondrites (−1.9±0.1 ε-unit) (Kleine et al., 2004), andε182WA is the initial Solar System value, and in practice, this term isusually set to be the initial value inferred from refractory inclusions(−3.47±0.20 ε-units) (Kleine et al., 2005). Model ages calculated usingEq. (2) are shown in Table 1 and Fig. 1. Most magmatic irons havemodel ages relative to CAIs that are zero within uncertainties, and

some samples have significantly negative model ages, confirmingprevious findings. The paradoxical negative Hf–Wmodel ages of someiron meteorites can have two possible causes. One is that the actualinitial ε182W value in CAIs is lower than −3.47±0.20, and the other isthat ε182W values measured in iron meteorites have been altered byexposure to GCR.

Palme et al. (1994) and Humayun et al. (2007) found evidence forW redistribution in Allende CAIs accompanying parent body meta-morphism and alteration of the carbonaceous chondrites. Palme et al.(1994) showed that W was depleted in opaque assemblages (OAs)from the Allende Egg 6 CAI. They also found complementary Wenrichments in surrounding silicates, suggestive of W redistributionduring formation of OAs. Humayun et al. (2007) reported a metal veincross-cutting a silicate mineral grain in the Allende Golfball CAI with avery high W concentration (N100 ppm) compared to surroundingareas, providing evidence for the transport of oxidized W in the CAI.According to Humayun et al. (2007), this could potentially affect theinterpretation of the 182Hf-182W isochron reported for Allende CAIs byKleine et al. (2005), since W in OAs might have isotopically exchangedwith radiogenic W in the silicates during metamorphism.

In contrast with the ambiguity in the initial ε182Wof CAIs, there arestrong lines of evidence that ε182W values in some iron meteoriteswere modified by exposure to GCR (Markowski et al., 2006b). A modelsimulation (Masarik, 1997) predicted a maximum shift of −0.5 inε182W for the Toluca iron meteorite (3.9-m radius and 600 Myexposure age). Thus GCR effect can explain some (if not all) of thenegative model ages. A remaining question is how to correct for theseeffects on metal-silicate differentiation ages.

4.1. The effect of galactic cosmic ray irradiation and correction of modelages

As a first approach towards quantifying the GCR effect on Wisotopic compositions, the ε182W values are plotted against exposureages for magmatic irons in Fig. 4. The exposure ages are from (Signerand Nier, 1962; Schultz, 1967; Nord and Zahringer, 1972; Voshage andFeldmann, 1979; Villa et al., 1981; Voshage et al., 1983; Voshage, 1984;Olsen et al., 1994; Nishiizumi et al., 1995; Lavielle et al., 1997; Lavielleet al., 1999), and are mostly 41K–40K ages. The 41K–40K ages are usuallymore precise than the exposure ages based on other chronometers butmay still suffer from a systematic bias. Nevertheless, it can be seen inFig. 4 that ε182W values tend to decrease with increasing exposureages. The samples with exposure ages of less than 400 My form a

Fig. 5. Schematic diagrams illustrating the method applied for correcting thecosmogenic effect on ε182W in magmatic iron meteorites. A) Cosmogenic effects onthe W isotopic compositions of four iron meteorites (# 1–4) of the same magmaticgroup. The original pre-exposure ε182W value of the group is assumed to be −3.47(a priori unknown). Exposure to GCR can only drive the ε182W values to the low side.The lengths of the solid arrows represent the actual shifts due to the cosmogenic effect.The solid circles represent the measured W isotopic compositions. B) The ε182W valuesof all four samples were corrected for the maximum cosmogenic effect (open arrows).The open symbols show the corrected W isotopic compositions. The minimum of thecorrected values (# 2) defines the upper limit and the maximum of the uncorrectedvalues (# 1) defines the lower bound of the pre-exposure ε182W value of this magmaticiron meteorite group.

99L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

plateau, defining an average ε182W value of −3.47±0.20, identical tothe initial value of refractory inclusions (Kleine et al., 2005). Magmaticiron meteorites from the same group are thought to sample the sameplanetesimal core, so they should record the same metal-silicatedifferentiation event and hence should have the same initial ε182W(prior to GCR exposure). The fact that ε182W is variable within a singlemagmatic group or sometimes within a single sample is most likelycaused by different exposures to GCR. One does not expect a simplecorrelation between ε182W values and exposure ages within eachgroup because the production rates of cosmogenic nuclides depend onmany factors besides exposure age, such as flux of primary andsecondary high-energy cosmic ray particles, the chemical compositionof the iron meteorite, and the sample location within the meteorite.Thus, it is impossible to correct for the cosmogenic effect in individualsamples using the exposure age information alone.

Markowski et al. (2006b) proposed a method for correcting theGCR effect using the concentration of another cosmogenic nuclide, 3He(Markowski et al., 2006b). They demonstrated a correlation of ε182Wvalue with cosmogenic 3He concentration for both Grant (IIIB) andCarbo (IID) meteorites. For samples not too far from the pre-atmospheric surface, a positive correlation between ε182W and 3Heconcentration is expected. They were thus able to extract the pre-exposure ε182W values based on their model results of ε182W variationwith 3He. We did not attempt to correct cosmogenic effects using 3Heconcentrations for Grant and other samples since the method hasseveral limitations. One is that meteorites are susceptible to 3He losson heating (Schultz, 1967). Also, different types of nuclear reactionsare involved in the production of cosmogenic 3He and W isotopes.Helium-3 is primarily produced by spallation processes throughinteractions of high-energy GCR particles with target elements such asO, Mg, Si, Ca, Fe and Al. As a result, the maximum 3He production rateoccurs at the pre-atmospheric surface of themeteorite. In contrast, theproduction/destruction of W isotopes occurs through capture ofsecondary thermal neutrons that have relatively low energy. Theproduction of thermal neutrons reaches a maximum at a certaindepth, as do the neutron capture reaction rates. As a result of thesefactors, the correlation between 3He and ε182W is not very robust.

Here, a new method was developed and applied to estimate thepre-exposure ε182W value for each magmatic iron meteorite group,which relies on the assumption that samples from the same group(same planetesimal core) have the same pre-exposure ε182W value.The method proceeds as follows (Fig. 5):

• For each group, the highest measured ε182W value is taken torepresent a lower bound on the initial ε182W. The reason is that GCRcan only shift ε182W towards lower values and the highest valuerepresents a minimum for the pre-exposure ε182W in the group.

• The ε182W values of all samples belonging to a group were correctedfor maximum possible cosmogenic effects. Inevitably, all ε182Wvalues must have been over corrected because we apply themaximum conceivable correction (i.e., for the optimum depth toyield the largest correction at a given exposure age and pre-atmospheric radius). Thus the least radiogenic value among thecorrected values must be an upper limit to the pre-exposure ε182Win the group.

Measurement uncertainties were taken into account whenestimating the values of the lower and upper limits. The maximumpossible cosmogenic effects in individual meteorites were computedby combining our model simulations of neutron capture rates on Wisotopes, estimated pre-atmospheric radii (when available) (Voshage,1984), and exposure ages of individualmeteorites (see the footnotes ofTable 1 for references). According to Masarik (1997), the maximumneutron capture rate on W isotopes, and hence the maximum deficitin ε182W, occurs at a certain depth beneath the pre-atmosphericsurface. The pre-atmospheric size of the meteorite may affect thedepth where the maximum effect takes place and also the magnitude

of the effect. The depth- and size-dependent production rates ofcosmogenic W were calculated by integrating the spectra of primaryand secondary particles with excitation functions of relevant nuclearreactions. The spectra of primary and secondary particles werecalculated using the LAHET code (Masarik and Reedy, 1994). Thedifferential particle spectra depend on the solar modulation para-meter M, which is a measure of solar activity and therefore varies withspace and time (Leya et al., 2000). The calculated spectra are forM=550 MeV (Leya et al., 2000). A GCR proton (N10 MeV) flux of2.86 cm−2

s−1 was adopted (Kollar et al., 2006). Three main libraries

(ENDF, JEF, and KASKAD) give cross sections for neutron inducedreactions relevant for the production of W isotopes and were used forcomparisonpurposes. The computed effects are not sensitive to Fe andNi concentrations. Fig. 6 shows a plot of maximum cosmogenic shift inε182W against the pre-atmospheric radius, assuming a constantexposure age of 657 My. For a given pre-atmospheric radius, themaximum possible effect on ε182W is linearly correlated withexposure age, thus the effect at any exposure age can be calculatedfrom Fig. 6. The magnitude of the modeled effect with the ENDFlibrary is about 1.52 and 1.89 times higher than those with the JEF andKASKAD libraries, respectively. The modeled results obtained withthe JEF library were chosen for correction because of the followingreasons:

(i) We computed depth profiles for Carbo (group IID; pre-atmospheric radius 0.53 m; exposure age 850 My) and Grant(group IIIB; 40 cm; 695 My) using the ENDF, JEF, and KASKADlibraries and compared these with measured profiles (Markowskiet al., 2006b, Table 2) in Fig. 2B and C; ENDF predicts a steeperslope than what is observed, while JEF and KASKAD give good fitsto the measurements.(ii) The modeled GCR effects with the ENDF library are unrealis-tically large, e.g., the maximum effect is up to 1.5 ε for a 650 Myexposure. If we assume that the original ε182W values of ironmeteorites are around −3.45, the maximum cosmogenic shiftwould have led to a present day value of ∼−5.0. As can be seen inFig. 4, such a negative value has not been measured in any ironmeteorite with high-precision methods.

Fig. 6. Model results of maximum cosmogenic effects on ε182W for meteorites withvarious pre-atmospheric radii and constant exposure age of 657 My.

Fig. 7. A) Metal-silicate differentiation model ages and B) accretion ages for magmaticirons relative to the time of CAI formation. The metal-silicate differentiation model ages

100 L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

(iii) Although JEF and KASKAD give equally good fits to themeasured profiles, the former library was preferred because theuncertainties on pre-exposure ε182W based on that library aremore conservative.

For samples with unknown pre-atmospheric size, the largestpossible correction was applied (corresponding to a pre-atmosphericradius of 0.7 m). Precise W isotope measurements reported in recentwork (Scherstén et al., 2006; Markowski et al., 2006a,b) were includedin our calculations and the ranges of pre-exposure ε182W values thuscalculated for each magmatic iron meteorite group are shown inTable 2 and Fig. 7; these are −3.63 to −3.35, −3.43 to −3.23, −3.95 to−3.24, −3.40 to −3.28, −3.41 to −3.33, −3.35 to −3.00, −3.43 to −3.28and −3.5 to −3.46 for IC, IIAB, IID, IIIAB, IIIE, IIIF, I VA and IVB irons,respectively. For IID irons, only two samples have been studied so far:Carbo and Hraschina. A value of −3.14±0.12 was reported forHraschina (Markowski et al., 2006a), thus indicating a very radiogeniclower bound. However, the depth profile of Carbo suggests a pre-exposure ε182W of ∼−3.65 (Fig. 2C). To be conservative, the leastnegative value of Carbo analyses was used as the lower bound (whichcould still have been affected significantly by GCR effect). The fact that

Table 2ε182W values in magmatic iron meteorites corrected for cosmogenic effects

Group ε182W ΔTM–S ΔT′M–S R ΔTaccretion ΔT′accretion

(My) (My) (km) (My) (My)

IC −3.63 to −3.35 −1.2 to 1.0 −3.0 to 2.6 52 0–0.7 0–1.6IIAB −3.43 to −3.23 0.3 to 2.1 −1.4 to 3.6 94 0.1–1.4 0–1.8IID −3.95 to −3.24 −3.4 to 2.0 −5.2 to 3.5 170 0–1.3 0–1.8IIIAB −3.40 to −3.28 0.6–1.6 −1.2 to 3.1 41–50 0.4–1.1 0–1.7IIIE −3.41 to −3.33 0.5–1.2 −1.2–2.7 120 0.3–0.8 0–1.6IIIF −3.35 to −3.00 1.1–4.6 −0.7 to 6.1 35 0.7–1.9 0–2.0IVA −3.43 to −3.28 0.3–1.6 −1.4 to 3.2 7–18 0.15–1.1 0–1.7IVB −3.50 to −3.46 −0.2 to 0 −2.0 to 1.6 3–21 0 0–1.1

ΔTM–S represents the model age of metal-silicate differentiation calculated with Eq. (2)and adopting an initial ε182W CAI value of −3.47. ΔT′M–S is a more conservative metal-silicate differentiation age, considering an uncertainty of ±0.2 ε in the initial value ofCAIs. R stands for the conventional value of the radius of iron meteorite parent bodycalculated from metallographic cooling rates, based on models without regolith (Haacket al., 1990). ΔTaccretion represents the accretion age of individual iron meteorite parentbodies derived from corresponding curves in Fig. 8 and using ΔTM–S. ΔT′accretion is amore conservative accretion time estimate (see text).

are calculated using Eq. (2) and the estimated ranges for ε182W values corrected forexposure to GCR. In 7B, the dark gray bars represent the preferred age estimates and thelight grey bars represent the conservative age estimates.

no single calculated upper bound is lower than the uncorrected lowerboundary suggests that this methodworks as intended. The estimatedranges of pre-exposure ε182W values for all groups overlap with theinitial CAI value. No single group shows pre-exposure ε182W that issignificantly more radiogenic than the CAI initial, except possibly theIIIF irons.

4.2. Time scales of metal-silicate differentiation

Metal-silicate differentiation model ages of magmatic groups arecalculated relative to the CAI initial of −3.47 (Table 2 and Fig. 7).Tight time constraints were obtained for the IC, IIAB, IIIAB, IIIE, IVAand IVB iron meteorite groups, indicating that core formation ontheir parent bodies very likely occurred within 1 to 2 My of theformation of CAIs. Not a single group shows a resolvable negativemodel age relative to CAIs. The IIIF group is the only one that stands

Fig. 8. Timing of metal-silicate differentiation vs. accretion age based on thermalmodeling of the evolution of asteroids with radii between 5.2 and 100 km. The numbersadjacent to the solid lines are the radii in kilometers. Given a differentiation age (from182Hf–182W systematics) and a parent body size (frommetallographic cooling rates), it ispossible to derive the accretion age (Table 2, Eq. (5)). Note that the adiabatic curve forincipient melting of Fe–FeS (computed assuming no heat loss, Eq. (8)) provides a robustupper limit on the accretion age given the differentiation age. The long dashed curverepresents adiabatic melting when the initial 60Fe/56Fe=5×10−7. For the range of metal-silicate differentiation ages obtained for most iron meteorite groups, the effect oninferred accretion time scales of adding such an amount of 60Fe is very small. Thedashed-dotted line is adiabatic melting with an Al content of 1.75%. All other curves aremodeled using an Al content of 1.13%. The dotted line represents the extreme case inwhich the time of differentiation is equal to that of accretion. The forbidden zonereflects the fact that after accretion, differentiation cannot proceed faster than in thecase of adiabatic heating.

101L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

out with metal-silicate differentiation possibly occurring as late as5 My after CAI formation.

4.3. Modeling the accretion time scales of iron meteorite parent bodies

As stated in the above section, all magmatic iron meteorite parentbodies underwent rapid metal-silicate differentiation. In the case ofmost magmatic groups, this event was predated by the condensationof refractory inclusions in the early Solar System by at most 5 My. TheHf–W model ages can be combined with thermal modeling toconstrain the accretion timescales of their parent bodies.

For a spherical planetesimal, the temperature T at any radialdistance R and at any time t after its formation is governed by the heatconduction equation:

ATAt

¼ 1R2

A

ARR2κ

ATAR

� �þ A r; tð Þ

Cð3Þ

where C is the specific heat capacity, A(r,t) is the heat production perunit mass, and κ is the thermal diffusivity (κ ¼ K

Cρ, where K is thethermal conductivity and ρ is the density). One effective and attestedheat source for the differentiation of early formed planetesimals is thedecay of short-lived radionuclides, such as 26Al (Lee et al., 1976) and60Fe (Shukolyukov and Lugmair, 1993; Tachibana and Huss, 2003;Mostefaoui et al., 2005; Tachibana et al., 2006). With an initial 26Al/27Al ratio of 5×10−5 (Macpherson et al., 1995; Russell et al., 1996; Husset al., 2001), a half-life of 0.73 My and a decay energy of 3 MeV/atom,26Al could have been an important heat source in the first few millionyears of the Solar System. Iron-60 (t1/2=1.49 My) is another possibleheat source but its initial abundance is under great debate. Giventhe inferred low and uncertain initial 60Fe/56Fe ratio (on the order of5×10−7, Tachibana and Huss, 2003; Mostefaoui et al., 2005; Tachibanaet al., 2006; Cook et al., 2006), it is considered separately. Thus, theterm A in Eq. (3) can be represented mathematically as

A r; tð Þ ¼ At0;Ale−λAlt ; ð4Þ

where λAl is the decay constant of 26Al. At0,AL =A0,Ale−λAlt

0 is the heatproduction by 26Al decay at the time of planetesimal formation, t0,where A0,A1 is the heat production at the time of CAI formation. κ, ρ,and C are all temperature-dependent (Yomogida and Matsui, 1983;Ghosh and McSween, 1999). Since we are only interested in the first-order correlation between accretion and differentiation timescales,appropriate constant values have been assigned to these parameters.In that case, Eq. (3) can be solved analytically (Carslaw and Jaeger,1959; Miyamoto et al., 1981):

T ¼ T0 þκAt0;Al

κλAle−λAltðR sinðr λAl=κð Þ1=2Þ

r sinðR λAl=κð Þ1=2Þ−1Þ

þ2R3At0 ;Al

rπ3K∑∞

n¼1

−1ð Þnn n2−λR2=κπ2ð Þ sin

nπrR

e−κn2π2t=R2

ð5Þ

The values chosen for κ and ρ in (LaTourrette and Wasserburg,1998; Hevey and Sanders, 2006) were used here. For C, a value of800 J/kg−1K−1 was used (Ghosh and McSween, 1999). T0 is the initialtemperature of the planetesimal, which is equal to the ambienttemperature. Eq. (5) applies to a system with a fixed boundarycondition, where the surface temperature is always equal to theambient temperature. The ambient temperature varies with time andheliocentric distance. A value of 250 K was adopted, which is the sameas that used in (Hevey and Sanders, 2006), and is within the range ofexpected temperatures (100–400 K) in the mid-plane of the solar diskat 2.5 AU and at 1 My after formation of the Sun (Woolum and Cassen,1999). The chemical compositions of iron meteorite parent bodies forlithophile elements like Al are uncertain. Previous studies have usedAl concentrations of both carbonaceous and ordinary chondrites for

modeling purposes. It is assumed here that the planetesimals have thecomposition of H-type ordinary chondrites, which was used in aprevious thermal model for the HED parent body (Ghosh andMcSween, 1998). Using an Al concentration of 1.13 wt.% (Wasson,1988), the calculated initial heat production for a planetesimal formedat t=0, is 1.9×10−7 W/kg. The model parameters are summarized inTable S1.

It is assumed that core formation is instantaneous and corre-sponds to the time when the center of the parent body reaches theeutectic temperature for the Fe–FeS system. This eutectic tempera-ture at 1 bar is 1261 K (Fei et al., 1997). Electrical conductivitymeasurements show that Fe–S alloy will segregate at temperatureshigh enough to melt the alloy (∼1261 K), but below the silicatesolidus (Yoshino et al., 2003). The segregation velocity is rapid incomparison with the timescale of core formation (Yoshino et al.,2003). The trigger of Fe–S melting is set at the center, because for allplanetesimals with radii N50 km, Fe–FeS melting at various depthstakes place at about the same time, except for an outermost shell of afew kilometers in thickness. The proportion of this “cold” shell to thewhole planetesimal radius increases with decreasing planetesimalsize.

Planetesimals accreted at different times have different initial heatproduction rates At0,Al. The time that it takes to initiate Fe–FeS melting(onset of core formation) can be calculated analytically using Eq. (5)(Fig. 8). For planetesimals accreted at the same time, the larger theradius is, the sooner the core forms. For planetesimals of b∼5.2 kmradius, melting of Fe–FeS will never occur, unless a regolith (with verylow thermal diffusivity) is present to reduce heat loss (Haack et al.,1990). A general feature of the melting-accretion relation in Fig. 8 isthat for each radius, there is a “critical accretion time”, after which theplanetesimals will not undergo metal-silicate differentiation. Thecritical accretion time is sensitive to the radius for smaller planeste-simals (Rb20 km). Planetesimals formed later than 2.4 My after CAIs

102 L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

will never undergo metal-silicate differentiation, regardless of theirsizes.

The calculations show that for large planetesimals of ≥50 kmradius, melting in the interior is adiabatic. In that case, the time that ittakes to initiate melting in the Fe–FeS system is determined by theenergy balance between radiogenic heat produced by decay of 26Aland change in internal energy,

A tð Þdt ¼ CdT ð6ÞIntegrating both sides of Eq. (6) from accretion to core formation, itfollows,

∫ tCt0 A0;Ale−λAltdt ¼ ∫TeutT0

CdT ; ð7Þ

where tC is the time of core formation. T0 and Teut are the initialtemperature and eutectic temperature of the Fe–FeS system. tC can bewritten as:

tC ¼ −1λln e−λt0−Cλ

Teut−T0ð ÞA0;Al

� �: ð8Þ

Using metallographic cooling rates, Chabot and Haack (2006)recalculated the radii of iron meteorite parent bodies for severalgroups of irons, when taking into account the presence of a regolithwith a thickness of 0.003 times the radius. The new calculated radii areapproximately 2-fold lower than conventional values obtained with-out a regolith (Haack et al., 1990). However, we did not incorporate aregolith in our thermal modeling and used conventional radii instead.The effect of this omission to the model results (Fig. 8) is discussedbelow.

For eachmagmatic group, the size of the parent body is known andthus the corresponding evolution curve can be found in Fig. 8. Withthe metal-silicate differentiation age inferred from 182Hf-182Wsystematics, constraints can be put on the accretion time, denotedas ΔTaccretion in Table 2. Our preferred accretion timescales are within∼1.5 My of the formation of CAIs for most magmatic iron groups.

Several sources of uncertainties for the accretion time need to beaddressed:

(i) The differentiation ages computed with the thermal modelcorrespond to incipient melting of Fe–FeS at the center of theplanetesimal (onset of core formation) while 182Hf–182Wmodelages correspond to large-scale metal-silicate differentiation.The calculated accretion ages based on Fig. 8 and 182Hf–182Wmodel ages must therefore represent upper limits.

(ii) In the thermal model, the possible existence of a regolith wasnot incorporated. Adding a thin regolith will decrease the timerequired to melt the metal and cause the curves in Fig. 8 to shiftto the right for small parent body sizes (towards the adiabaticlimit). This will increase the accretion age for a given metal-silicate differentiation age and a given parent body radius.However, the radius of the parent body derived from metallo-graphic cooling rates with a thin regolith is smaller than thevalue used here (computed without a regolith). This will shiftthe accretion time in the opposite direction. For large parentbodies, adding a regolith will not affect the model results at allsince the interiors will melt adiabatically. The bottom line hereis that the adiabatic curve represents a robust upper limit forthe accretion times.

(iii) Based on the diverse cooling rates that have been measured forIVA iron meteorites (Goldstein and Short, 1967; Rasmussen,1982; Yang et al., 2007), Yang et al. (2007) inferred a largerparent body size of 300-km diameter. Thus, melting in the IVAparent body would be closer to adiabatic melting.

In brief, when model uncertainties are considered, the accretiontime inferred from the adiabatic curve gives a robust upper limit for

any given 182Hf–182Wmetal-silicate differentiation age and any parentbody radius. A conservative accretion time, denoted as ΔT′accretion inTable 2, was calculated taking into account both the thermal modeluncertainties and an uncertainty of ±0.2 ε in the initial ε182W value ofthe CAIs (−3.47). The accretion times of all magmatic groups are, forthemost part, within 1 to 2My after the formation of the Solar System.

It should be noted that the model did not take into account 60Fe asa heat source. The abundance of 60Fe in the planetesimal will affect thetime of melting. If the initial 60Fe/56Fe ratio is about 5×10−7, theadiabatic curve will shift towards the right (Fig. 8). The latest accretiontime for meteorite parent bodies of any size that would undergometal-silicate differentiation is ∼2.4 My (Fig. 8). For the range ofmetal-silicate differentiation ages reported for most iron meteoritegroups, the effect on inferred accretion time scales of adding such anamount of 60Fe is very small. The Al content in iron meteorite parentbodies is also uncertain. The highest value of 1.75 wt.% among allchondrite groups (Wasson, 1988) shifts the adiabatic curve to theright, corresponding to a critical accretion time of ∼2.5 My (Fig. 8).

Our results show that accretion of iron meteorite parent bodiesoccurred very early, within 1–2 My of CAI formation. Although theabsolute accretion rate of planetesimals in the Solar System is not verywell constrained, studies have shown that the accretion timescaleincreases with increasing heliocentric distance (Weidenschilling,1977; Greenberg et al., 1978; Grimm and McSween, 1993). Ironmeteorite parent bodies are usually assumed to have formed in themain asteroid belt. However, Bottke et al. (2006) showed thatdifferentiated meteorite parent bodies may not be indigenous to themain asteroid belt but may be interlopers from the terrestrial planetformation region, where fast accretion rates allowed small planete-simals to melt early. The very early accretion ages of most magmaticiron meteorite parent bodies implied by the fast core differentiationtimes are consistent with this scenario.

4.4. Clues to the origin of non-magmatic iron meteorites

The origins of non-magmatic groups IAB–IIICD and IIE are underdebate and it is not clear whether or not samples from the same groupshould possess a single ε182W value. IABs and IIICDs belong to a singlegroup of non-magmatic meteorites (Wasson and Kallemeyn, 2002).Three out of four analyses of IAB–IIICD ironmeteorites studied presenthomogeneous ε182W values of −3.01 to −2.80 (Table 1; Fig. 1). This isconsistent with previous studies (Scherstén et al., 2006; Markowskiet al., 2006a). Markowski et al. found that some IIICDs have resolvablylower values (∼−3.3) than IABs (∼−3.0) and concluded that IIICD ironsformed earlier (contemporaneous with CAIs) than IABs. The only IIICsample from this study, Nantan, yields a ε182W value of −2.93±0.07,which is identical with those of the two IABs. It is clear that its Wisotopic composition is not affected by the cosmogenic effect given itsshort exposure age and extremely heavy shielding (Nishiizumi et al.,1995). However Nantan is one of the most weathered iron meteoritesand terrestrial contamination associated with weathering cannot beexcluded, even though the sample was intensively leached beforedissolution. Deport presents the most unradiogenic ε182W (−3.82)among all IAB–IIICDs, which clearly resulted from exposure to GCRirradiation (given its long exposure age of 1140 Ma). Wasson andKallemeyn (2002) distinguished six subgroups of IAB–IIICD irons.They suggested that melting occurred because of impact heating andthat the slight compositional difference between themain group (MG)and low-Au subgroups (SLL, SLM, SLH) resulted from independentimpact events that occurred either at different locations of a singlechondritic body or different chondritic bodies with similar composi-tions (Wasson and Kallemeyn, 2002). Most of the IAB–IIICD samplesstudied for W isotopes are from the main group (MG) and thesubgroup SLL. The similarity of W isotope compositions of IAB andIIICD irons is most easily explained with a single impact event dated at4–7 My after CAI formation causing multiple metal pools in a single

103L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

chondritic parent body. Alternatively, there may have been multipleimpacts occurring within a narrow time interval of 4–7 My after CAIformation on multiple chondritic bodies with similar chemicalcompositions.

For the other non-magmatic group IIE, Watson has the mostradiogenic signature (ε182W=−2.30±0.08) among all iron meteoritesmeasured thus far (also see Snyder et al., 2001). The other sampleArlington has a much lower ε182W of −3.01±0.08. With an exposureage of only 150 My, it is unlikely to have been affected by cosmogeniceffects (Lavielle et al., 1999). IIE iron meteorites also contain abundantsilicate inclusions. The heterogeneity in ε182W values of IIEs isconsistent with a formation model that involves melting and metalsegregation followed by partial mixing of metal and silicate by asecondary thermal event (Bogard et al., 2000). Thus, as has beenadvocated for stony-iron meteorites by (Quitté et al., 2005), the Wisotopic compositions of IIEs may result from isotope exchangebetween metal and radiogenic silicate during mixing of the two andare unlikely to have any chronological meaning. In order to betterconstrain their genetic history, more IIE iron meteorite samples willneed to be analyzed.

5. Conclusions

The present study demonstrates that exposure to galactic cosmicrays can create strong negative shifts in ε182W values by capture ofsecondary thermal neutrons. This agrees with most recent studies(Kleine et al., 2005; Scherstén et al., 2006; Markowski et al., 2006a,b).After correction of these effects, the ε182W values in various magmaticgroups (except IID and IIIF) are confined to a narrow range of −3.63 to−3.23. This implies that core segregation processes in most magmaticironmeteorite parent bodies occurred b2Myafter the formation of theSolar System. This result also agrees with earlier studies (Schersténet al., 2006; Markowski et al., 2006a). However, IIIF iron meteoritestend to have more radiogenic ε182W compositions, which may imply adelayed metal-silicate differentiation event, up to 5 My after CAIformation. Metal-silicate differentiation ages from 182Hf–182W sys-tematics (Table 1) were combined with parent body sizes (frommetallographic cooling rates) into a model of planetesimal heating by26Al-decay (Fig. 8) in order to constrain the accretion timescales of ironmeteorite parent bodies (Table 2, Fig. 7). The early differentiation agesare consistent with heating by 26Al-decay and rapid accretion (within1.5 My of CAI formation). The fast accretion timescales are consistentwith dynamical models suggesting that the parent bodies of ironmeteorites were formed in the inner region of the protoplanetary disk,where accretion rates were presumably high (Weidenschilling, 1977;Greenberg et al., 1978; Grimm and McSween, 1993). The non-magmatic IAB–IIICD group has a homogeneous ε182W isotopiccomposition. This is best explained by melting of a chondritic objectby a single late impact event that occurred 4–7 My after the formationof CAIs. The variable ε182W signatures of IIEs are consistent with aformation model that advocates melting and metal segregationfollowed by partialmixing ofmetal and silicate by a secondary thermalevent. A key uncertainty that remains in all 182Hf–182Wage estimates isthe initial ε182W of CAIs (Kleine et al., 2005; Humayun et al., 2007).

Acknowledgement

We thank the Field Museum and the Smithsonian Institution forproviding iron meteorite samples. We are grateful to Timothy McCoyfor his guidance in the selection of the Grant specimen. L. Qinacknowledges support in the form of a graduate fellowship from theField Museum. L. Qin is grateful to Robert N. Clayton and LawrenceGrossman for their helpful reviews on her dissertation thesis relatedto this manuscript, and to C. Nicole Foley for generously sharing thedetails of her W separation protocol. We also thank Eric Gilabert forgiving us access to his compilation of meteorite exposure ages. We

thank H. Palme and K. Mezger for their insightful reviews of themanuscript. This work was supported by the France-Chicago Center, aPackard Fellowship and NASA grants NNG06GG75G (to N. D.) andNNG05GG22G (to M. W.).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.epsl.2008.06.018.

References

Ammon, K., Masarik, J., Leya, I., 2006. Cosmogenic production rates in iron meteorites.Meteorit. Planet. Sci. 41, A16.

Bogard, D.D., Garrison, D.H., McCoy, T.J., 2000. Chronology and petrology of silicatesfrom IIE iron meteorites: evidence of a complex parent body evolution. Geochim.Cosmochim. Acta 64, 2133–2154.

Bottke, W.F., Nesvorný, D., Grimm, R.E., Morbidelli, A., O'Brien, D.P., 2006. Ironmeteorites as remnants of planetesimals formed in the terrestrial planet region.Nature 439, 821–824.

Carlson, R.W., Hauri, E.H., 2001. Extending the 107Pd–107Ag chronometer to low Pd/Agmeteorites with multicollector plasma-ionization mass spectrometry. Geochim.Cosmochim. Acta 65, 1839–1848.

Carslaw, H.S., Jaeger, J.C., 1959. Conduction of Heat in Solids. Clarendon Press, Oxford.510 pp.

Chabot, N.L., Haack, H., 2006. Evolution of asteroidal cores. In: Lauretta, D.S., McSweenJr.Jr., H.Y. (Eds.), Meteorites and the Early Solar System, vol. II. The University ofArizona, Tucson, AZ, pp. 747–771.

Chen, J.H., Wasserburg, G.J., 1996. Extinct 107Pd in the early solar system andimplications for planetary evolution. In: Basu, A., Hart, S. (Eds.), Earth Processes:Reading the Isotopic Code, Geophysical Monograph, vol. 95. AGU, Washington, DC,pp. 1–20.

Cook, D.L., Wadhwa, M., Janney, P.E., Dauphas, N., Clayton, R.N., Davis, A.M., 2006. Highprecision measurements of non-mass-dependent effects in nickel isotopes inmeteoritic metal via multicollector ICPMS. Anal. Chem. 78, 8477–8484.

Fei, Y., Bertka, C.M., Finger, L.W., 1997. High-pressure iron-sulfur compound, Fe3S2, andmelting relations in the Fe–FeS system. Science 275, 1621–1623.

Foley, C.N., Wadhwa, M., Borg, L.E., Janney, P.E., Hines, R., Grove, T.L., 2005. The earlydifferentiation history of Mars from 182W–142Nd isotope systematics in the SNCmeteorites. Geochim. Cosmochim. Acta 69, 4557–4571.

Ghosh, A., McSween, H.Y., 1998. A thermal model for the differentiation of Asteroid 4Vesta, based on radiogenic heating. Icarus 134, 187–206.

Ghosh, A., McSween, H.Y., 1999. Temperature dependence of specific heat capacity andits effect on asteroid thermal models. Meteorit. Planet. Sci. 34, 121–127.

Goldstein, J.I., Short, J.M., 1967. Cooling rates of 27 iron and stony-iron meteorites.Geochim. Cosmochim. Acta 31, 1001–1023.

Greenberg, R., Wacker, J.F., Hartmann, W.K., Chapman, C.R., 1978. Planetesimals toplanets—numerical-simulation of collisional evolution. Icarus 35, 1–26.

Grimm, R.E., McSween, H.Y., 1993. Heliocentric zoning of the asteroid belt by 26Alheating. Science 259, 653–655.

Haack, H., Rasmussen, K.L., Warren, P.H., 1990. Effects of regolith megaregolith insulationon the coolinghistories ofdifferentiatedasteroids. J.Geophys. Res.– SolidEarth Planets95, 5111–5124.

Hevey, P.J., Sanders, I.S., 2006. A model for planetesimal meltdown by 26Al and itsimplications for meteorite parent bodies. Meteorit. Planet. Sci. 41, 95–106.

Hoffman, J.H., Nier, A.O., 1958. Production of helium in iron meteorites by the action ofcosmic rays. Phys. Rev. 112, 2112–2117.

Horan, M.F., Smoliar, M.I., Walker, R.J., 1998. 182W and 187Re–187Os systematics of ironmeteorites: chronology for melting, differentiation, and crystallization in asteroids.Geochim. Cosmochim. Acta 62, 545–554.

Humayun, M., Simon, S.B., Grossman, L., 2007. Tungsten and hafnium distribution incalcium–aluminum inclusions (CAIs) from Allende and Efremovka. Geochim.Cosmochim. Acta 71, 4609–4627.

Huss, G.R., MacPherson, G.J., Wasserburg, G.J., Russell, S.S., Srinivasan, G., 2001.Aluminum-26 in calcium–aluminum-rich inclusions and chondrules from unequi-librated ordinary chondrites. Meteorit. Planet. Sci. 36, 975–997.

Jacobsen, S.B., Harper, C.L., 1996. Accretion and early differentiation history of the Earthbased on extinct radionuclides. In: Basu, A., Hart, S. (Eds.), Earth Processes: Readingthe Isotope Code, Geophysical Monogram, vol. 95. AGU, Washington, DC, pp. 47–74.

Kleine, T., Münker, C., Mezger, K., Palme, H., 2002. Rapid accretion and early coreformation on asteroids and the terrestrial planets from Hf–W chronometry. Nature418, 952–955.

Kleine, T., Mezger, K., Münker, C., Palme, H., Bischoff, A., 2004. 182Hf–182W isotopesystematics of chondrites, eucrites, and Martian meteorites: chronology of coreformation and early mantle differentiation in Vesta and Mars. Geochim.Cosmochim. Acta 68, 2935–2946.

Kleine, T., Mezger, K., Palme, H., Scherer, E., Münker, C., 2005. Early core formation inasteroids and late accretion of chondrite parent bodies: evidence from 182Hf–182Win CAIs, metal-rich chondrites, and iron meteorites. Geochim. Cosmochim. Acta 69,5805–5818.

Kollar, D., Michel, R., Masarik, J., 2006. Monte Carlo simulation of GCR neutron captureproduction of cosmogenic nuclides in stony meteorites and lunar surface. Meteorit.Planet. Sci. 41, 375–389.

104 L. Qin et al. / Earth and Planetary Science Letters 273 (2008) 94–104

LaTourrette, T., Wasserburg, G.J., 1998. Mg diffusion in anorthite: implications for theformation of early solar system planetesimals. Earth Planet. Sci. Lett. 158, 91–108.

Lavielle, B., Remazeilles, C., Marti, K., 1997. Solar helium and neon in the Arlington IIEiron meteorite. Meteorit. Planet. Sci. 32, A78.

Lavielle, B., Gilabert, E., Marti, K., 1999. Records of solar energetic particles in the ironmeteorite Arlington. Meteorit. Planet. Sci. 34, A72–A73.

Lee, D.C., 2005. Protracted core formation in asteroids: evidence from high precision Wisotopic data. Earth Planet. Sci. Lett. 237, 21–32.

Lee, D.C., Halliday, A.N., 1995. Precise determinations of the isotopic compositions andatomicweights ofmolybdenum, tellurium, tin and tungsten using ICPmagnetic sectormultiple collector mass spectrometry. Int. J. Mass Spectrom. Ion Process. 146/147,35–46.

Lee, D.C., Halliday, A.N., 1996. Hf–W isotopic evidence for rapid accretion anddifferentiation in the early solar system. Science 274, 1876–1879.

Lee, T., Papanastassiou, D.A., Wasserburg, G.J., 1976. Demonstration of 26Mg excess inAllende and evidence for 26Al. Geophys. Res. Lett. 3, 41–44.

Leya, I., Lange, H.J., Neumann, S., Wieler, R., Michel, R., 2000. The production ofcosmogenic nuclides in stony meteoroids by galactic cosmic-ray particles. Meteorit.Planet. Sci. 35, 259–286.

Leya, I., Wieler, R., Halliday, A.N., 2003. The influence of cosmic-ray production onextinct nuclide systems. Geochim. Cosmochim. Acta 67, 529–541.

Macpherson, G.J., Davis, A.M., Zinner, E.K., 1995. The distribution of 26Al in the earlySolar-System—a reappraisal. Meteoritics 30, 365–386.

Maréchal, C.N., Télouk, P., Albarède, F., 1999. Precise analysis of copper and zinc isotopiccompositions by plasma-source mass spectrometry. Chem. Geol. 156, 251–273.

Markowski, A., Quitté, G., Halliday, A.N., Kleine, T., 2006a. Tungsten isotopiccompositions of iron meteorites: chronological constraints vs. cosmogenic effects.Earth Planet. Sci. Lett. 242, 1–15.

Markowski, A., Leya, I., Quitté, G., Ammon, K., Halliday, A.N., Wieler, R., 2006b.Correlated helium-3 and tungsten isotopes in iron meteorites: quantitativecosmogenic corrections and planetesimal formation times. Earth Planet. Sci. Lett.250, 104–115.

Masarik, J., 1997. Contribution of neutron-capture reactions to observed tungstenisotopic ratios. Earth Planet. Sci. Lett. 152, 181–185.

Masarik, J., Reedy, R.C., 1994. Effects of bulk composition on nuclide productionprocesses in meteorites. Geochim. Cosmochim. Acta 58, 5307–5317.

Miyamoto, M., Fujii, N., Takeda, H., 1981. Ordinary chondrite parent body: An internalheating model. Proc. Lunar Planet. Sci. 12B, 1145–1152.

Mostefaoui, S., Lugmair, G.W., Hoppe, P., 2005. 60Fe: a heat source for planetarydifferentiation from a nearby supernova explosion. Astrophys. J. 625, 271–277.

Nishiizumi, K., Finkel, R.C., Caffee, M.W., 1995. Cosmogenic radionuclides in Donglingand Nantan iron-meteorites—case of heavily shielded Chinese twins. Meteoritics30, 556–557.

Nord, R., Zahringer, J., 1972. Cosmogenic helium, neon and argon in 33 iron meteorites.Contribution to Recent Geochemistry and Analytical Chemistry, Nauka, Moscow,pp. 59–71.

Olsen, E., Davis, A., Clarke, R.S., Schultz, L., Weber, H.W., Clayton, R., Mayeda, T.,Jarosewich, E., Sylvester, P., Grossman, L., Wang, M.S., Lipschutz, M.E., Steele, I.M.,Schwade, J., 1994. Watson—a new link in the IIE iron chain. Meteoritics 29, 200–213.

Palme, H., Hutcheon, I.D., Spettel, B., 1994. Composition and origin of refractory-metal-rich assemblages in a Ca, Al-rich Allende inclusion. Geochim. Cosmochim. Acta 58,495–513.

Qin, L., Dauphas, N., Janney, P.E., Wadhwa, M., 2007. Analytical developments for high-precision measurements of W isotopes in iron meteorites. Anal. Chem. 79,3148–3154.

Qin, L., Dauphas, N., Janney, P.E., Wadhwa, M., 2008. Tungsten nuclear anomalies inplanetesimal cores. Astrophys. J. 674, 1234–1241.

Quitté, G., Birck, J.L., 2004. Tungsten isotopes in eucrites revisited and the initial 182Hf/180Hf of the solar system based on iron meteorite data. Earth Planet. Sci. Lett. 219,201–207.

Quitté, G., Birck, J.L., Allègre, C.J., 2005. Stony-iron meteorites: History of the metalphase according to tungsten isotopes. Geochim. Cosmochim. Acta 69, 1321–1332.

Rasmussen, K.L., 1982. Determination of the cooling rates and nucleation histories of 8group IVA iron-meteorites using local bulk Ni and P variation. Icarus 52, 444–453.

Russell, S.S., Srinivasan, G., Huss, G.R., Wasserburg, G.J., MacPherson, G.J., 1996. Evidencefor widespread 26Al in the solar nebula and constraints for nebula time scales.Science 273, 757–762.

Scherstén, A., Elliott, T., Hawkesworth, C., Russell, S., Masarik, J., 2006. Hf–W evidencefor rapid differentiation of iron meteorite parent bodies. Earth Planet. Sci. Lett. 241,530–542.

Schoenberg, R., Kamber, B.S., Collerson, K.D., Eugster, O., 2002. NewW-isotope evidencefor rapid terrestrial accretion and very early core formation. Geochim. Cosmochim.Acta 66, 3151–3160.

Schultz, L., 1967. Tritium loss in iron meteorites. Earth Planet. Sci. Lett. 2, 87–89.Shen, J.J., Papanastassiou, D.A., Wasserburg, G.J., 1996. Precise Re–Os determinations

and systematics of iron meteorites. Geochim. Cosmochim. Acta 60, 2887–2900.Shukolyukov, A., Lugmair, G.W., 1993. Live iron-60 in the early Solar System. Science

259, 1138–1142.Signer, P., Nier, A.O., 1960. The distribution of cosmic-ray-produced rare gases in iron

meteorites. J. Geophys. Res. 65, 2947–2964.Signer, P., Nier, A.O., 1962. The measurement and interpretation of rare gas

concentrations in iron meteorites. In: Moore, C.B. (Ed.), Researches on Meteorites.Wiley, New York, pp. 7–35.

Smoliar, M.I., Walker, R.J., Morgan, J.W., 1996. Re–Os ages of group IIA, IIIA, IVA, and IVBiron meteorites. Science 271, 1099–1102.

Snyder, G.A., Lee, D.C., Ruzicka, A.M., Prinz, M., Taylor, L.A., Halliday, A.N., 2001. Hf–W,Sm–Nd, and Rb–Sr isotopic evidence of late impact fractionation and mixing ofsilicates on iron meteorite parent bodies. Earth Planet. Sci. Lett. 186, 311–324.

Sugiura, N., Hoshino, H., 2003. Mn–Cr chronology of five IIIAB ironmeteorites. Meteorit.Planet. Sci. 38, 117–143.

Tachibana, S., Huss, G.R., 2003. The initial abundance of 60Fe in the Solar System.Astrophys. J. 588, L41–L44.

Tachibana, S., Huss, G.R., Kita, N.T., Shimoda, G., Morishita, Y., 2006. 60Fe in chondrites:debris from a nearby supernova in the early Solar System. Astrophys. J. 639,L87–L90.

Villa, I.M., Huneke, J.C., Wasserburg, G.J., 1981. Spallogenic rare-gases in iron-meteoriteswith anomalous silver. Earth Planet. Sci. Lett. 56, 9–18.

Vockenhuber, C., Oberli, F., Bichler, M., Ahmad, I., Quitté, G., Meier, M., Halliday, A.N., Lee,D.C., Kutschera, W., Steier, P., Gehrke, R.J., Helmer, R.G., 2004. New half-lifemeasurement of 182Hf: improved chronometer for the early solar system. Phys. Rev.Lett. 93, #172501.

Völkening, J., Köppe, M., Heumann, K.G., 1991. Tungsten isotope ratio determinations bynegative thermal ionization mass-spectrometry. Int. J. Mass Spectrom. Ion Process.107, 361–368.

Voshage, H., 1984. Investigations of cosmic-ray-produced nuclides in iron-meteorites, 6.The Signer–Nier model and the history of the cosmic-radiation. Earth Planet. Sci.Lett. 71, 181–194.

Voshage, H., Feldmann, H., 1979. Investigations on cosmic-ray-produced nuclides iniron-meteorites, 3. Exposure ages, meteoroid sizes and sample depths determinedby mass-spectrometric analyses of potassium and rare-gases. Earth Planet. Sci. Lett.45, 293–308.

Voshage, H., Feldmann, H., Braun, O., 1983. Investigations of cosmic-ray-producednuclides in iron-meteorites, 5. More data on the nuclides of potassium and noble-gases, on exposure ages and meteoroid sizes. Z. Naturforsch. 38a, 273–280.

Wasson, J.T., 1988. Compositions of chondrites. Philos. Trans. R. Soc. Lond., A 325,535–544.

Wasson, J.T., Kallemeyn, G.W., 2002. The IAB iron-meteorite complex: a group, fivesubgroups, numerous grouplets, closely related, mainly formed by crystalsegregation in rapidly cooling melts. Geochim. Cosmochim. Acta 66, 2445–2473.

Weidenschilling, S.J., 1977. Distribution of mass planetary system and solar nebula.Astrophys. Space Sci. 51, 153–158.

Woolum, D.S., Cassen, P., 1999. Astronomical constraints on nebular temperatures:implications for planetesimal formation. Meteorit. Planet. Sci. 34, 897–907.

Yang, J.J., Goldstein, J.I., Scott, E.R.D., 2007. Iron meteorite evidence for early formationand catastrophic disruption of protoplanets. Nature 446, 888–891.

Yin, Q.Z., Jacobsen, S.B., Yamashita, K., Blichert-Toft, J., Télouk, P., Albarède, F., 2002. Ashort timescale for terrestrial planet formation from Hf–W chronometry ofmeteorites. Nature 418, 949–952.

Yomogida, K., Matsui, T., 1983. Physical-properties of ordinary chondrites. J. Geophys.Res. 88, 9513–9533.

Yoshino, T., Walter, M.J., Katsura, T., 2003. Core formation in planetesimals triggered bypermeable flow. Nature 422, 154–157.